The proliferation of deepfake technology, leveraging deep learning algorithms to manipulate facial features, attributes, and expressions in images, has elicited significant apprehension. Consequently, a burgeoning body of research aims at identifying images synthesized by deepfake algorithms. Although Vision Transformer-based methods have showcased commendable performance in image recognition, recent investigations suggest a decline in deepfake detection compared to convolutional neural network-based techniques. This study, proposes a high-precision deepfake detection approach employing the Wavelet Vision Transformer, incorporating self-supervised learning. The Wavelet Vision Transformer demonstrates proficiency in capturing essential high-frequency components within images, particularly pertinent for deepfake detection. By amalgamating it with self-supervised learning, a variant of representation learning, our method facilitates the precise detection of manipulation artifacts within deepfake images, thereby attaining elevated detection accuracy.

We propose a method to enhance accuracy by improving the pixel selection method in the circuit configuration of template matching using limited pixels. By selecting a pair of pixels with a low co-occurrence probability or a large difference in pixel value from the left and right pixels at a certain distance from the pixel selected based on image features, matching accuracy can be improved while using the same circuit as the conventional one.

Micro-hole drilling with a drill of 0.1 mm or less requires constant monitoring of machining conditions to prevent tool breakage. The micro-hole drilling machine previously developed by the authors' group is equipped with sensors to detect drilling torque and thrust force and can display both drilling forces in a real-time graph on the machine operation panel. However, the machine did not have the function to output data externally, and thus the data could not be analyzed in detail. In this study, we developed a machining force data processing device that converts signals from the sensors into numerical data at high speed and outputs the data wirelessly. And the resolution of the torque sensor was improved to enable detection of minute machining forces in drilling holes with a drill diameter of 0.03 mm. And we confirmed that the sensors had sufficient sensitivity by comparing the data detected by the sensor with the dynamometer value. Then, a calibration value was calculated using the comparison value.

Previous work has clarified that the waving amplitude of a linear motion ball guide (LMBG) can be reduced by optimizing for the shape of crowning on each end of raceways in a carriage. Waving in a table using LMBG becomes in 1/100 μm order amplitude by the optimization; it is in 1/10 μm order without the optimization. Moreover, the waving analysis method considering table deformation and misalignment of assembly has also been developed. In recent years, waving has been required in 1/1000 μm order. However, it is analytically confirmed that only crowning configuration optimization does not achieve this request. Considering the principle of waving generation, increasing balls in LMBG is the most effective to suppress the waving. Although balls should become small not to change the carriage length and to increase balls, balls becoming smaller degrades the load capacity of LMBG. Balls should be drastically increased further to keep capacity and suppress waving. For the problem, a new LMBG design with the number of raceways to eight by double-rowing has been led; ordinary LMBG has four raceways. This paper reports the development process and evaluation results of the LMBG with the new design, although the product has already been launched.

In recent years, due to the progress of machine learning, the classification accuracy of normal and anomaly classes by image processing has been improved. There are many situations where ”the environment of imaging for learning classifiers” and ”the environment of imaging in the inspection site” are different in the general manufacturing site. In particular, it is assumed that cameras may differ due to malfunctions or model changes. Images obtained from two different cameras may have the same visual appearance, however, differences in sensor and sensitivity cause differences in each pixel. This difference affects the classifier that judges between normal and anomaly using machine learning and hinders the achievement of accurate automatic visual inspection. To solve this problem, we propose an image conversion method that reduces the differences between two cameras. The image obtained by the camera for inspection is translated to the image obtained by the camera for learning of classifier. In the evaluation experiment, the proposed method achieved an accuracy of 97.0% in classifying normal and anomaly images. The proposed method improved accuracy by 3.8% compared to the previous method without image conversion (93.2%).

This paper presents the effects of setting and operating conditions on the performance of an oil-air spindle. While several studies have explored the dynamic stiffness of running spindles, few have focused on analyzing these spindles to their operational states. This paper introduces a method for electromagnetic excitation and dynamic compliance measurement to characterize the dynamic behavior of an oil-air spindle in response to an applied excitation force. The dynamic compliance of the spindle was measured under varying conditions, including spindle speed, bearing temperature, spindle orientation (vertical or horizontal), and lubrication supply parameters. The impact of spindle setting and operating conditions on dynamic performance was analyzed with reference to eigen frequencies and damping ratios. The analysis revealed two significant spindle bending modes that exhibit distinct changes in response to spindle speed and temperature. Additional experiments demonstrated slight shifts in eigen frequency and variations in damping ratio concerning the oil supply.

This study deals with a reduction of burr height on edge of micromilled channels. The microchannels were cut in oxygen-free copper by using cemented carbide end mill under semidry condition. The oxygen-free copper with superior conductivity is an electrode material for EDM, but it is a material difficult to cut. The microchannels of 50μm depth were cut by multistep slotting to reduce burr height. Axial depth of each slotting was set so that total depth became 50μm. Each slotting was carried out in recommended cutting conditions. The micro burr height was measured by SEM. The mechanism of burr formation at the edge of the micromilled channel is similar to entry burr formation in drilling. In drilling, entry board is often used to prevent entry burr formation. In multistep slotting, the edge of mere several micron thickness which made by preceded slotting acted as entry board in following slotting. This study shows the multistep slotting was effective in reducing the burr height by setting appropriate combination of axial depths.

In high-precision mechanical systems, such as machine tools, thermal deformation significantly impacts accuracy. This paper proposes a design method for meta-mechanical structures to reduce effective thermal deformation in mechanical systems using topology optimization. The proposed design method consists of two steps. In the first step, the configuration of Material 1 is optimized to maximize stiffness. In the second step, while keeping the configuration of Material 1 fixed, the configuration of Material 2, which has a different thermal expansion coefficient, is optimized to reduce thermal deformation. This study aims to minimize relative position changes and angular changes in the target area to control effective thermal deformation. The validity of the proposed method was verified by applying it to a three-dimensional design problem.